Share your thoughts, 1 month free Claude Pro on usSee more

3D Robotic Manipulation on PrimitiveSkill via ManiSkill

100Place Success Rate

GPT-5

Updated 15d ago

Evaluation Results

Method	Links
GPT-5 2026.05		100	63	0	100	55
o3 2026.05		100	63	0	100	82
o4-mini 2026.05		100	50	0	75	33
Qwen2.5-VL-72B 2026.05		100	50	0	100	33
VAGEN 2026.05		100	88	88	88	50
AtlasVA 2026.05		100	100	100	100	100
Gemini 2.5 flash 2026.05		75	50	0	88	82
Gemini 2.0 2026.05		75	13	0	25	33
Gemini 2.5 Pro 2026.05		63	63	0	75	0
Claude Sonnet 4.5 2026.05		63	50	0	100	100
Claude Sonnet 3.7 2026.05		63	13	0	100	90
GPT-4o 2026.05		50	63	0	88	94
Qwen2.5-VL-7B 2026.05		0	0	0	75	3
Qwen2.5-VL-3B 2026.05		0	0	0	0	0
VLM-R1-3B 2026.05		0	0	0	0	0